Model comparison
DeepSeek-V3 vs Mistral Medium
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 36.3 on the Noometry Index.
Last verified . 29 shared benchmarks.
DeepSeek
Mistral AISummary
- They share 29 benchmarks with published results for both. DeepSeek-V3 scores higher in 3 categories and Mistral Medium in 5 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3 leads 37.5 to 25.0.
- The biggest single-benchmark swing is Vectara Hallucination Rate: 6.1% for DeepSeek-V3 and 22.7% for Mistral Medium.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $1.50 / $7.50 for Mistral Medium.
- Mistral Medium accepts more context: 262K tokens versus 164K.
Side by side
| DeepSeek-V3 | Mistral Medium | |
|---|---|---|
| Provider | DeepSeek | Mistral AI |
| Noometry Index | 39.5 | 36.3 |
| Released | 2024-12-26 | 2023-12-11 |
| Weights | Open | Open |
| Context window | 164K | 262K |
| Max output | 164K | 262K |
| Input $ / M tokens | $0.24 | $1.50 |
| Output $ / M tokens | $0.90 | $7.50 |
| Results tracked | 60 | 36 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
DeepSeek-V3: 42.3 (#106), Mistral Medium: 34.2 (#243)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| SciCode | 35.8% | 40.2% |
| WeirdML | 36.1% | 43.7% |
| LMArena Coding | 1368 | 1434 |
| FrontierCode | — | 8% |
| Aider Polyglot | 55.1% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 763.98 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Mistral Medium: 28.3 (#90)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 37.7% |
| METR Time Horizons | 49.6% | — |
Reasoning Mistral Medium leads
DeepSeek-V3: 20.5 (#236), Mistral Medium: 24.0 (#167)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 50% |
| CritPt | 0% | 0% |
| LMArena Hard Prompts | 1365 | 1426 |
| DTBench | 64.8% | 75.5% |
| LMCA | 15.5% | 26.1% |
| SimpleBench | 27.2% | — |
| LiveBench Reasoning | 65.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| Surface Evolver Bench | — | 26.9% |
| BIG-Bench Hard | 87.5% | — |
| Epoch Capabilities Index | 135.94 | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math DeepSeek-V3 leads
DeepSeek-V3: 32.1 (#219), Mistral Medium: 28.1 (#245)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 32.2% |
| LMArena Math | 1373 | 1408 |
| MATH Level 5 | 75.5% | 81.6% |
| FrontierMath (Feb 2025 set) | 1.7% | 0.3% |
| ProofBench | — | 9% |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
Knowledge DeepSeek-V3 leads
DeepSeek-V3: 37.5 (#155), Mistral Medium: 25.0 (#265)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| GPQA Diamond | 67.6% | 59.5% |
| Vectara Hallucination Rate | 6.1% | 22.7% |
| LMArena Expert | 1351 | 1408 |
| Humanity's Last Exam | — | 4.5% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multimodal Not comparable
DeepSeek-V3: —, Mistral Medium: 35.3 (#88)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| LMArena Vision | — | 1172 |
Multilingual Mistral Medium leads
DeepSeek-V3: 48.5 (#143), Mistral Medium: 52.1 (#91)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| LMArena Non-English | 1358 | 1408 |
| LMArena Chinese | 1391 | 1447 |
| LMArena French | 1385 | 1459 |
| LMArena German | 1374 | 1432 |
| LMArena Japanese | 1333 | 1378 |
| LMArena Korean | 1319 | 1380 |
| LMArena Russian | 1373 | 1411 |
| LMArena Spanish | 1358 | 1433 |
Instruction Following Too close to call
DeepSeek-V3: 72.8 (#130), Mistral Medium: 73.7 (#116)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| LMArena Instruction Following | 1345 | 1398 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |
Long Context Mistral Medium leads
DeepSeek-V3: 34.0 (#253), Mistral Medium: 42.9 (#114)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| LMArena Longer Query | 1352 | 1406 |
| Fiction.LiveBench | 50% | — |
Writing & Preference Mistral Medium leads
DeepSeek-V3: 57.4 (#130), Mistral Medium: 60.0 (#103)
| Benchmark | DeepSeek-V3 | Mistral Medium |
|---|---|---|
| LMArena Text | 1375 | 1424 |
| LMArena Creative Writing | 1364 | 1391 |
| Short-Story Creative Writing | 77% | 77.3% |
| LMArena Multi-Turn | 1389 | 1418 |
| EQ-Bench Creative Writing | 1472 | — |
| WildBench | 83% | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than Mistral Medium?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 36.3 on the Noometry Index.
Which is cheaper, DeepSeek-V3 or Mistral Medium?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Mistral Medium lists at $1.50 and $7.50.
Is DeepSeek-V3 or Mistral Medium better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 versus 34.2 in the Noometry coding category.
Which has the bigger context window?
Mistral Medium does, with 262K tokens against 164K.
How many benchmarks do DeepSeek-V3 and Mistral Medium share?
29 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Mistral Medium has 36.